Planning for repository disk and cluster multicast IP address

Edit online

PowerHA® SystemMirror® clustering can be done through multicast or unicast networking. To choose the multicast mode of clustering, you can plan and provide a multicast IP address for communication within the cluster. By default, if no multicast IP address is provided while deploying the cluster, PowerHA deploys unicast (normal TCP/IP socket communication) based cluster.

Cluster repository disk

You must have one active repository disk per cluster for standard clusters and stretched clusters. You can identify up to six backup repository disks per cluster for standard clusters and stretched clusters. You must have one active repository disk per site for linked clusters. You can identify up to six backup repository disks per site for linked clusters.

PowerHA SystemMirror uses a shared disk to store Cluster Aware AIX® (CAA) cluster configuration information. You must have at least 512 MB and no more than 460 GB of disk space allocated for the cluster repository disk. This configuration is automatically kept highly available on the disk that is provided. This feature requires that a dedicated shared disk be available to all nodes that are part of the cluster. This disk cannot be used for application storage or any other purpose.

When planning the disks that you want to use as repository disks, you must plan for a backup or replacement disks, which can be used in case the primary repository disk fails. The backup disk must be the same size and type as the primary disk, but could be in a different physical storage disk. Update your administrative procedures and documentation with the backup disk information. You can also replace a working repository disk with a new one to increase the size or to change to a different storage subsystem. To replace a repository disk, you can use the SMIT interface.

Note: If the shared disk that is used as a repository disk is a mapped virtual SCSI (vSCSI) disk, you must map the disk as an vSCSI disk to all nodes in the cluster. The mapping of the vSCSI disk must be identical across all nodes in the cluster. For example, you cannot map the repository disk using the vSCSI method to one node in the cluster and map the same disk using the N-Port ID Virtualization (NPIV) method to another node in the cluster.

Cluster multicast IP address

You can use a multicast IP address for cluster monitoring and communication. You can specify this address when you create the cluster, or you can have one be generated automatically when you synchronize the initial cluster configuration.

Note: The default mechanism uses unicast communications and requires no extra configuration. However, if you want to use multicast communication, you must continue reading and ensure that your network devices are enabled for multicast communications.

If you decide to use multicast, PowerHA SystemMirror uses multicast-based communication between hosts in the cluster. Your environment’s network must allow multicast IP packets to flow between hosts in the cluster. To verify whether nodes in your environment support multicast based communication, use the mping command. Run the mping command before you start using PowerHA SystemMirror in your environment.

Note: Some of the network switches allow multicast packets to flow for a while before stopping them. So, it is critical to conduct the mping test for at least 5 minutes and to make sure that the network fabric allows multicast packet flow without any issues. Also, when switches are cascaded, typically the switches need additional configuration to route multicast packets. To configure multicast packet flow, see the documentation that is provided by the switch vendor to configure multicast packet flow.

A multicast address is also known as a class D address. Every IP datagram whose destination address starts with 1110 is an IP multicast datagram. The remaining 28 bits identify the multicast group on which the datagram is sent. You must configure your kernel to receive packets sent to a specific multicast group, which makes the host join the group in the interface you specified.

Do not use the following multicast groups:

224.0.0.1: This is the all-hosts group. If you ping that group, all multicast-capable hosts on the network should answer, because every multicast-capable host must join that group at startup on all its multicast-capable interfaces.
224.0.0.2: This the all-routers group. All multicast routers must join that group on all its multicast capable interfaces.
224.0.0.4: This is the all DVMRP routers.
224.0.0.5: This is all OSPF routers.
224.0.013: This is all PIM routers.

Note: The range 224.0.0.0-224.0.0.255 is reserved for local purposes, such as administrative and maintenance tasks, and data that they receive is never forwarded by multicast routers. Similarly, the range 239.0.0.0 - 239.255.255.255 is reserved for administrative scoping. These special multicast groups are regularly published in the Assigned Numbers RFC.

PowerHA SystemMirror 7.1.2, or later, supports IP version 6 (IPv6), however, you cannot explicitly specify the IPv6 multicast address. CAA uses an IPv6 multicast address which is derived from the IP version 4 (IPv4) multicast address. To determine the IPv6 multicast address, a standard prefix of oxFF05 is combined using the logical OR operator with the hexadecimal equivalent of the IPv4 address. For example, the IPv4 multicast address is 228.8.16.129 or 0xE4081081. The transformation by the logical OR operation with the standard prefix is 0xFF05:: | 0xE4081081. Thus, the resulting IPv6 multicast address is 0xFF05::E408:1081.